propensities and counterfactuals: the loser that almost won 0lli... · attitudes and social...

ATTITUDES AND SOCIAL COGNITION

Propensities and Counterfactuals: The Loser That Almost Won

Daniel Kahneman and Carol A. VareyUniversity of California, Berkeley

Close Counterfactuals are alternatives to reality that "almost happened." A psychological analysisof close Counterfactuals offers insights into the underlying representation of causal episodes andthe inherent uncertainty attributed to many causal systems. The perception and representation ofcausal episodes is organized around possible focal outcomes, evoking a schema of causal forcescompeting over time. We introduce a distinction between two kinds of assessments of outcomeprobability: dispositions, based on causal information available prior to the episode; and propensi-ties, based on event cues obtained from the episode itself. The distinction is critical to the use ofalmost, which requires the attribution of a strong propensity to the counterfactual outcome. Thefinal discussion focuses on characteristic differences between psychological and philosophicalapproaches to the analysis of Counterfactuals, causation, and probability.

The question of how people think of things that could havehappened but did not has attracted increasing interest amongpsychologists in recent years (J. T. Johnson, 1986; Kahneman &Miller, 1986; Kahneman & Tversky, 1982a; Landman, 1987;D. T. Miller, Turnbull, & MacFarland, 1990; Wells & Gavanski,1989; Wells, Taylor, & Turtle, 1987). As philosophers have longknown, the study of Counterfactuals cannot be separated from aconception of causality, and an understanding of causality re-quires a conception of possibility and conditional probability.Counterfactual assertions rest on causal beliefs, and causal at-tributions invoke counterfactual beliefs, for example, aboutwhat would have happened in the absence of a putative cause.Some counterfactual assertions assign degrees of probability orplausibility to unrealized outcomes, many causal beliefs areprobabilistic, and judgments of probability often draw on im-pressions of causal tendencies or propensities. The present arti-cle is concerned with a psychological analysis of this nexus ofissues.

Our study began with an attempt to understand the psychol-ogy of assertions of the form "X almost happened," which wecall close Counterfactuals. An important characteristic of such

This article is based on a talk presented at the Conference on Coun-terfactual Thinking and Other Forms of Mental Simulation, BrownCounty, Indiana, April 1989. The research was supported by the AirForce Office of Scientific Research, Air Force Systems Command,U.S. Air Force, under Grant AFOSR-88-0206 to Daniel Kahneman.

We are grateful to Maya Bar-Hillel, Baruch Fischhoff, Dale Griffin,Lenore Kahneman, Igal Kvart, Dale Miller, and Amos Tversky forhelpful comments and conversations; to Anne Treisman for her in-sights throughout the project; and to three anonymous reviewers foruseful suggestions.

Correspondence concerning this article should be addressed toDaniel Kahneman, Department of Psychology, Tolman Hall, Univer-sity of California, Berkeley, California 94720.

assertions is that they are not expressed as a conditional with aspecified antecedent, as counterfactual conditionals are. Theclose counterfactual does not invoke an alternative possibleworld, but states a fact about the history of this world—namelythat things were close to turning out differently than they did.

Our approach combines some elementary phenomenologi-cal observations and an equally elementary linguistic inquiryinto the conditions under which close counterfactual assertionsare appropriate. The genre is not unknown in psychology:Heider (1958) and Schank and Abelson (1977), in particular,have successfully carried out ambitious exercises in this vein.Studies of what people mean when they say that "John went tothe restaurant" or when they use the words can and try havecontributed significantly to an understanding of how peoplethink about events and actions. In this article we examine theuse of the word almost in a speculative attempt to explore howpeople think about Counterfactuals, probability, and causa-tion.1 The present analysis is restricted to cases in which "Xalmost happened" implies that X could have happened. Weignore figurative uses of almost in which it is used to denote"coming close" without implication of possibility, as in "at thatbend the train almost touches the embankment." We also re-strict our discussion of almost to cases in which either the ac-tual outcome or the close counterfactual is an achievement (seeLyons, 1977; G. A. Miller & Johnson-Laird, 1976; Vendler,1967)—a change of state that occurs at a particular moment,usually as the culmination of a longer causal episode. We ana-lyze the beliefs that a speaker expresses by the assertion that an

1 The importance of the word almost as an indication of cognitivelyand emotionally relevant alternatives to outcomes that actually mate-rialized was pointed out by Heider (1958, pp. 141-144), who drewattention to an instructive passage in Henry Fielding's Tom Jones(1749/1975, Vol. 2, p. 691); see also Hofstadter(1979, pp. 634-643).

Journal of Personality and Social Psychology, 1990, Vfal. 59, No. 6,1101-1110Copyright 1990 by the American Psychological Association, Inc. 0022-3514/90/S00.75

1101

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

1102 DANIEL KAHNEMAN AND CAROL A. VAREY

individual almost died, or almost missed a deadline, and exam-ine what such beliefs imply to us about the cognitive representa-tion of uncertain events and of causal propensities.

This article develops the following ideas: (a) Counterfactuals,causes, and (some) probabilities are treated as facts about theworld, not as constructions of the mind, (b) The absence ofperfect hindsight indicates that people attribute inherent un-certainty to causal systems—what happened is not treated asnecessary or inevitable, (c) The perception and representationof causal episodes is organized around possible outcomes of theepisode, (d) Probabilities of outcomes can be assessed on thebasis of advance knowledge (dispositions) or of cues gainedfrom the causal episode itself (propensities). The distinction iscritical to the use of almost, which requires the attribution of astrong propensity to the counterfactual outcome, (e) Cues topropensity are the temporal or causal proximity of the focaloutcome and indications of rapid progress through a causalscript, (f) A general schema of causal forces competing overtime is applicable to many achievement contexts, (g) There arecharacteristic differences between a psychological and a philo-sophical approach to the analysis of probability, causality, andcounterfactuals.

The Counterfactual Stance

The statement "Xalmost happened" implies several ancillarybeliefs. It commits the speaker to the belief that another ob-server with the same information would agree with the counter-factual assertion, as would be the case with public, objectivefacts. The close counterfactual also implies that X could havehappened, denying the necessity or inevitability of what actu-ally happened and implicitly denying the deterministic charac-ter of the situation. We examine these beliefs in the followingsections.

Objective Reference

By definition, counterfactual statements refer to events thatdid not, in fact, occur. However, there is a compelling intuitionthat some counterfactuals are treated as having an objectivecharacter, not as mere mental constructions. Refuting the possi-ble-worlds analysis of counterfactuals, Goodman (1983) put thepoint strongly: "We have come to think of the actual as oneamong many possible worlds. We need to repaint that picture.All possible worlds lie within the actual one" (p. 57). In ascrib-ing objective status to counterfactuals, we intend to contrast theattitude toward these objects of thought from the attitude to-ward imaginings, fantasies, and desires, which are normallytagged as subjective (M. K. Johnson, 1988). The discriminationof what actually happened from what almost did is, of course,essential in the monitoring of reality—the counterfactual eventis not perceived as real, but it is not treated as subjective. The"fact" is that the outcome truly is close, or is not close, indepen-dently of anyone's beliefs.

The distinction between beliefs that have objective or subjec-tive status recalls an earlier discussion of alternative cognitiverepresentations of probability (Kahneman & Tversky, 1982b).Two main interpretations of that notion were identified, which

respectively assign it subjective or objective status. In the subjec-tive interpretation, which is standard in Bayesian philosophy, aprobability judgment describes the subject's degree of belief ina proposition. "The probability that the Nile is longer than theAmazon is p" is usually understood as describing the speaker'sbeliefs, not as a fact about these rivers. The attribution of sub-jective status to a belief is often marked by the possessive pro-noun: "My probability that the Nile is longer than the Amazon. . ." is more natural than "the probability that the Nile.. . "Incontrast, the probability that a thumbtack will land on its pointif tossed and the probability that Team A will defeat Team Bare normally understood as descriptions of the causal disposi-tions of the thumbtack or of the competing teams. A speakerwho wishes to indicate a subjective interpretation of probabilitywill use the possessive pronoun: "My probability that Team Awill win is. . ." acknowledges the possibility of valid alterna-tives, a stance that is not usually adopted in factual statements.

The two types of representation of uncertainty are mostclearly distinguished when the uncertainty is removed. The pos-sessive pronoun is then obligatory if the probability has subjec-tive status. "The probability that the Nile is longer than theAmazon was. . ."is simply anomalous. In contrast, it is reason-able to maintain that the probability that the thumbtack wouldland on its point was .55 even when it is known that it did notdo so on a particular instance. More interestingly, the statementthat "the probability that Team A would win was high. . ." isacceptable even if that team is known to have lost. The state-ment of past probability need not refer to anyone's beliefs at thetime of the episode. The statement could be made, for example,by a speaker who learned, after the game had ended, that aplayer of Team B had undertaken to throw the game if he hadan opportunity to do so. As this example illustrates, currentknowledge of the relevant causal factors may allow a speaker tosay retrospectively that an event that did not take place had highprobability—and as a special and rather extreme case to assertthat the event almost occurred.

Although in this article we often appeal to the reader's intu-itions in the expectation that they match ours, we also testedsome of our conclusions by collecting judgments of appropriate-ness from native speakers of English. Subjects were recruited onthe Berkeley campus by a poster offering students a small pay-ment in return for a completed questionnaire. Respondentswere given instructions and several questions as illustrated bythe examples below.

In the following questions you are asked to rate statements on ascale from "appropriate" to "very peculiar." One or more state-ments are presented for each question. Vbu are to rate whether thestatement in italics is appropriate, given the information in therest of the question.

\1. Tom almost died but in fact he was never in real danger.

/ Appropriate 7% Somewhat peculiar 27% Very peculiar 66%

(«=29)2

2 The number in parentheses refers to the number of respondentsanswering the question. In later examples, results will only be reportedfor the two extreme categories of response.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

PROPENSITIES AND COUNTERFACTUALS 1103

2. Everyone thought Phil almost died but in fact he was never inreal danger. r\/

Appropriate 69% Very peculiar 10% (n = 29)

3. The autopsy showed that when he was a child, Sid had sufferedfrom a rare childhood disease. The pathologist said that if thedisease had lasted a few days longer, it would have killed him. Noone knew about it at the time; they thought he had a mild case ofmeasles. Sid almost died as a child from that rare disease. n


These examples illustrate that the close counterfactual hasthe status of a historical fact. As is generally true when suchfacts are asserted, everything known to the speaker at the timeof the utterance can be relevant, but the beliefs of observers ofthe actual event are not. As shown by Example 2, the objectivestatus of close counterfactuals allows them to be believed erro-neously. Indeed, counterfactuals can be faked. Professionalwrestlers on television have perfected the art of appearing al-most to kill one another, but they avoided regulation by demon-strating that their occupation is actually quite safe.

Like counterfactuals and (some) probabilities, causal attribu-tions are also treated as objective facts about the world. This istrue of causality directly observed, as in the perception of acollision and in Michotte's (1946) demonstrations of launching.It is also true of the more abstract causes that are judged to raisethe (objective) probabilities of events or, in some contexts,render them inevitable (Mackie, 1974). The counterfactual as-sertion that an effect would not have occurred in the absence ofthe cause, the sine qua non condition of necessity, has the sameobjective character.

Inherent Uncertainty

The frequent mentions of counterfactual possibilities in ev-eryday discourse demonstrate a prevailing intuition that thingscould have been different, and in some cases almost were. Thisintuition commits the speaker to a particular set of beliefsabout causality. Specifically, X is neither necessary nor inevita-ble if it can properly be said that Y almost happened instead ofit. Naive intuitions are evidently not dominated by a pervasivebelief in strict determinism. Kvart (1986) reached a similar con-clusion in his discussion of counterfactual conditionals.

There is an intriguing tension between the intuition thatthings could have been otherwise and the well-known hindsighteffect, in which the inevitability of events that actually tookplace tends to be exaggerated. The evidence is compelling thatretrospective assessments of the probability of events are af-fected by knowledge of whether or not these events have takenplace (Fischhoff, 1975,1982). The term creeping determinismhas been used in this context. Two distinct forms of hindsighteffects are associated, respectively, with subjective and objectiveinterpretations of probability. The most common test of hind-sight effects requires the retrieval of a past state of belief: "Whatwas your probability at the end of 1988 that the Berlin Wallwould be opened within a year?" A hindsight bias is revealed insuch questions by a tendency to exaggerate the past subjectiveprobabilities of whatever is now known to be true. An exampleof an objective hindsight question could be "In the light ofcurrent knowledge, what was the probability in 1988 that the

Berlin Wall would be opened within a year?" A discrepancybetween prospective and retrospective probabilities is typicallyobserved in tests of objective as well as subjective hindsight(Fischhoff, 1975).

Unlike the subjective case, hindsight with an objective inter-pretation of probability is not necessarily a mistake. It is en-tirely reasonable for an observer to make inferences about acausal system from the knowledge that it produced a particularoutcome. Indeed, what is most puzzling in this context is thelimited extent of creeping determinism in retrospective evalua-tions of outcomes.

Of course, not all causal systems are uncertain. As illustratedby most people's attitudes toward the mechanical and elec-tronic devices that surround them, a belief in strict determin-ism does not require much understanding of how the systemworks; it cannot be ignorance about the causal system that pre-cludes determinism about close counterfactuals. It is an impor-tant fact about causal reasoning that a sense of the necessity ofconsequences is often absent. In particular, there is no sense ofnecessity or inevitability in considering games of chance, manycontests and competitions, some physical systems (e.g., weatherand chance devices), or intentional actions.

The Representation of Causal Episodes

The idea that perceived goals serve to organize the representa-tion of action and imbue events with meaning was articulatedby Heider (1958), and is at the core of the more recent treat-ments of scripts and story grammars (Black & Bower, 1979;Kintsch & van Dijk, 1978; Rumelhart, 1977; Schank, 1975;Trabasso, Secco, & van den Broek, 1984) and treatments ofdecision making and causal reasoning that rely on a story-basedaccount (Pennington & Hastie, 1988; Read, 1987). Our concep-tion of causal episodes generalizes this idea to achievements.The class of achievements contains the outcomes of intentionalaction but is much richer: Dying from a disease, a river over-flowing its bank, and the Dow-Jones index rising 1,000 pointsin a year are all achievements. Many achievements are asso-ciated with particular causal scripts. The representation of anepisode as an instantiation of a causal script is therefore orga-nized in terms of its possible focal outcomes, and attention todifferent achievements will alter the representation. The stormthat could fill the reservoirs could also ruin the cherry crop,and its representation will be different if attention is directed toone of these outcomes rather than the other.

To illustrate the function of focal outcomes, we introduce athought experiment to which we shall repeatedly return. Imag-ine observing a sequence of red and blue balls as they are drawnfrom an urn, or the representation of such a process on a com-puter screen. Note the potent effects of an intention to watch fora particular outcome, such as the color that is most frequentafter 11 draws, or an excess of four red balls or six blueballs—whichever happens first. Although these focal outcomesare not goals, they serve the same function in organizing theimpression of the sequence. Most important, watching thesame sequence with different outcomes in mind alters the expe-rience. Although our thought experiment involves real-time ob-servation and uncertainty, neither of these elements is essential:A designated outcome will affect the interpretation of a story,

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


and the effect is not reduced when thinking of an episodewhose outcome is already known.

The probability of the focal outcome may fluctuate in thecourse of a causal episode. Changes of probability are alwaysinvolved in close counterfactuals: Perhaps the most compellingintuition about the statement "X almost happened" is that theprobability of X must have been quite high at some point beforeit dropped—all the way to zero if another outcome eventuallyterminated the episode.3

There are several reasons for probability changing in thecourse of an episode. We turn again to the urn example toillustrate two types of probability change. If there was initialuncertainty about the composition of the urn, beliefs about theurn will change to accommodate observed events—by Bayes'srule for an ideal observer—and the probability of the focal out-come will change accordingly. In addition, the actual probabil-ity of the focal outcome also changes becauseof the interveningevents. Every red ball drawn makes it more probable that theaggregate outcome will be an excess of red over blue balls. Theprobability of the focal outcomes will change, more or less regu-larly and perhaps with large fluctuations, until a decisive eventbrings about an outcome that terminates the episode. Note thatthis situation can support a close counterfactual: It is easy toimagine a sequence of draws of which it can appropriately besaid that the focal outcome almost occurred (red almost won),or almost did not.

The same types of changes of probability will also be foundin observing (or hearing about) a storm that could cause a floodor a couple deciding on a joint future. The events that constitutethe episode reveal the strength of an underlying causal process,and also contribute to bring about or retard the outcome. Theyalso indicate possible changes in the causal system—changesthat could be modeled by an urn whose composition is modi-fied after each draw, perhaps in response to the draw.

Propensities and Dispositions

The discussion so far has been in terms of "objective" proba-bilities—in the chance example these are probabilities thatcould be computed precisely, given some initial beliefs aboutthe composition of the urn. It is evident from this example thatan account of almost in terms of probability has some appeal:The probability of the counterfactual outcome must have beenhigh at some point. It turns out, however, that an account thatrelies exclusively on objective probabilities will not work. Someaspects of the puzzle to be solved are illustrated by the follow-ing examples:

4. Mark tried to register for the chess tournament. Because of aproblem in mailing the form he missed the registration deadlineby one day. Mark is a much stronger player than all the partici-pants in the tournament. Mark almost won the tournament.


5. At the end of a long game of chance, John could have won thewhole pot if a die that he rolled showed a six. The die that he rolledwas loaded to show six 80% of the time. John rolled it and itshowed a two. The die almost showed six.


6. At the end of a long game of chance, John could have won thewhole pot if a die that he rolled showed a six. The die thalhe rolledwas loaded to show six 80% of the time. John rolled it and itshowed a two. John almost won the whole pot.


The close counterfactual is decisively rejected in Examples 4and 5 but not in Example 6, although the prior probability ofthe focal outcome was high in all cases. Before it was rolled, theprobability of the die showing six was .80 in Examples 5 and 6,and the prior probability of Mark winning the tournament wasalso high, though unspecified. Despite this, the intuition thatalmost is inappropriate in the first two examples is so strongthat they seem almost absurd. A strong belief in the counterfac-tual conditional "Mark would have won if he had played" is notsufficient to support the close counterfactual "Mark almostwon," even if it is also accepted that he almost played in thetournament. Why is this the case? And what else is required forthe close counterfactual to be appropriate?

The answer to the first question is that the close counterfac-tual is never appropriate if it is only supported by indications oflikelihood or causal force that were available before the onset ofthe relevant causal episode. Achievements, such as winning atournament, getting married, or a die showing six, are asso-ciated with causal scripts that usually have a definite startingpoint: when play begins, when the couple start dating, when thedie is rolled. Probabilities can be assigned to possible outcomesof a causal process before it is initiated: Mark may be a ratedplayer, the couple could appear severely mismatched, the diecould be loaded. We shall refer to the cognitive representationof such prior probabilities as the (perceived) disposition of acausal system to yield particular outcomes. Examples 4 and 5show that dispositions, however strong, do not suffice to sup-port the assertion of a close counterfactual.

A close counterfactual must be supported by the evidence ofevent cues, as these accumulate in the course of the causal epi-sode. We use the term propensity for what is learned about theprobability of an outcome from observing event cues or fromhearing about them. Mark had a disposition to win his chesstournament and probably would have won it if he had regis-tered, but the causal episode for his victory never began, andthere was therefore no opportunity to establish a propensity forthat outcome. The standard example of propensity in a chanceevent is the cinematic cliche of the roulette wheel that slowsdown as it approaches a critical number, slows down even more,leans against the spring, then finally trips it and stops on aneighboring number. To be described as almost showing six, adie must display a propensity to stop its roll in that position.

The contrasting responses to Examples 5 and 6 illustrate theneed to distinguish propensity from probability. We supposethat our respondents would have assigned a probability of .8both to the die showing six and to John winning the whole pot.However, the propensities of the two outcomes clearly differ.

3 Kvart (1986) has offered a treatment for a broad class of counterfac-tual conditionals in which causality is explicated by probabilities thatchange over time.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Example 6 illustrates a common structure in which one achieve-ment (the die showing six) is nested inside another (John win-ning the pot). Because the focal outcome of winning the potinvokes a more inclusive causal episode that had begun longbefore the critical play, John can be said to have had a propen-sity to win, even if it is not established that the die had a propen-sity to show six. Thus, although the two statements have thesame probability before the throw, the differential effects ofpropensity and disposition allow "John almost won" to be ap-propriate although "the die almost showed six" is not.

We should now review the rather subtle relations among theconcepts of disposition, propensity, and probability, as they areused in this article. Disposition has been denned as the cogni-tive representation of the probability of a focal outcome, beforethe beginning of the relevant causal episode. A disposition canbe assessed either prospectively or in hindsight, depending onwhether or not the outcome is known. Disposition is a psycho-logical construct, not a logical or mathematical one, and inview of what is known about intuitive judgment there is littlereason to expect dispositions to obey the standard axioms ofprobability (Kahneman, Slovic, & Tversky, 1982). Dispositionsrepresent knowledge about the particular causal system thatwill (or will not) produce the focal outcome of current concern.Dispositions are inferred from the base rates of outcomespreviously produced by that system (Mark has won most of histournaments) or from structural knowledge that supportscausal inferences (the die was loaded in a particular fashion).Thus, the concept of disposition has causal as well as statisticalimplications.

Our concept of propensity is even more imbued with causalcontent. Event cues reveal the causal system in action. Theyindicate advance toward the focal outcome, or regression awayfrom it. They suggest changes in the momentary state of thecausal system—changes that may be real or illusory, as when aplayer is seen to have a "hot hand" (Gilovich, Vallone, &Tversky, 1985). Perhaps most important, propensities dependon the proximity of the outcome, on the possibility of quicklyachieving a decisive advance to it. In sharp contrast to probabil-ity, the propensities for all competing outcomes of a processmay be low early in a causal episode, and more than one pro-pensity can be high at once when the end is close. These ideasare elaborated in subsequent sections.

Our main interest in the remainder of this article is to useclose counterfactuals to learn about propensity. We considerpropensity to be a dimension of the experience and cognitiverepresentation of events, just as pitch is a dimension of auditoryexperience. There should be no presuppositions about the deter-minants of propensity; in particular, propensity could reflectcausality as well as probability, just as pitch depends on boththe frequency and the intensity of sound. To anchor this specula-tive analysis in observables, we assume that the appropriatenessof almost, in its literal meaning, provides a usable indication ofhigh propensity.

Disposition Neglect

Dispositions and propensities are differentially susceptibleto revision in hindsight. Consider two cases in which the ob-

server of the last lap of a footrace might assign a high probabil-ity of victory to a particular runner: (a) a runner who is incontention and is known to have a strong finish, or (b) a runnerwho has been catching up rapidly with the leader. The real-timeexpectations are equally strong in both cases, we assume, butthey are based on different cues—dispositional knowledge in(a) and event cues in (b). Now imagine that the two runners bothfail to win, by the same amount: The first did not show a strongfinish and the second never quite caught up. Note that it will notdo to say of the runner who usually has a strong finish that he orshe almost won the race with a strong finish, if in fact he or sheshowed no evidence of talent on that particular occasion. Theclose counterfactual that the loser almost won is more applica-ble to (a) than to (b), although a counterfactual conditionalcould be appropriate in (a). The general hypothesis is that dis-positional expectations that are not confirmed by event cuesbecome irrelevant in hindsight.

The differential weighting of event cues and dispositionalexpectations in retrospective judgments will be called disposi-tion neglect; the effect bears an intriguing resemblance to therelative neglect of base-rate information that has been observedin some prospective judgments. For example, the judged proba-bility that a short personality sketch describes a lawyer ratherthan an engineer is not much affected by the proportion ofengineers and lawyers in the sample from which it was drawn.The information about the individual case largely supersedesthe information about the base rate instead of combining withit according to Hayes's rule (Kahneman & Tversky, 1973). Simi-larly, Ajzen (1977) found that people predicting exam successfor a student based their predictions on a descriptive sketch andgave little weight to the information that the student was drawnat random from a set selected by a researcher to include 75%failures. As Ajzen observed, however, the neglected base rate inthese examples is merely statistical. There is no causal connec-tion between the composition of the student sample and thefactors that would make a particular student succeed or fail.The situation changes when such a causal connection is pro-vided: The information that 75% of students taking the testfailed it leads readily to the inference that the test was a diffi-cult one, and the information has much more impact on thejudgment of the probable success of an individual (Ajzen,1977). There have been other demonstrations of the generalprinciple that causally relevant base-rate information will notbe neglected (Gigerenzer, Hell, & Blank, 1988; Tversky & Kah-neman, 1980,1982; see also Bar-Hillel, 1990, for a discussion ofthese issues).

Ajzen's (1977) experiment demonstrated that dispositionalinformation tends to dominate statistical base rates, and thatdispositional information from two sources (the difficulty ofthe exam and the student's ability) tends to be integrated. Avariation of this experiment would demonstrate disposition ne-glect: Evidence that a student is extremely able does not supportthe inference that the student almost passed an exam that hefailed, nor does the knowledge that a test was very hard supportthe conclusion that a student who passed almost failed it. Inassessing close counterfactuals, event cues dominate causalbase rates and other dispositional information.

The neglect of statistical base rates leads to violations of

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Bayes's rule in prospective judgments. The neglect of disposi-tional expectations in hindsight is not necessarily an error, butthe psychology of the two effects may well reflect a single gen-eral principle. In both cases the data that bear most directly onthe causal forces at work in the individual case have the greatestimpact.

Correlates of Propensity

In this section we develop the concept of propensity by exam-ining two of its close correlates: shrinking distance and increas-ing impact. The role of distance and motion in the close coun-terfactual is evident in the near synonymy of "X almost hap-pened," "^nearly happened," and "X was close to happening."These expressions invoke a rich metaphor in which an extendedcausal process is represented as movement in space (e.g., see the"source-path-goal" kinesthetic image schema analyzed in La-koff, 1987; the various "journey" metaphors in Lakoff & John-son, 1980; and the force and space images in Talmy, 1981,1983).This metaphor imposes a metric of causal distance betweensituations and suggests the closest approach to an outcome as ameasure of its propensity. The second correlate of increasingpropensity is an escalation in the apparent causal significanceof events as the outcome is approached.

Causal Proximity

The present analysis has emphasized causal processes thatextend over time, but close counterfactuals can be asserted onthe basis of a measure of proximity or similarity even when theprocess is instantaneous. For example, the statement "Thehouse was almost struck by lightning" is appropriate whenlightning struck nearby. The actual outcome is the only eventcue in such cases, and it induces a gradient of propensity in itsspatial and temporal vicinity. From the fact that lightningstruck in a particular place at a particular time, a propensity isinferred to strike in neighboring places, and at about the sametime. Similarly, it is appropriate to say that Tom almost got sixsixes in rolling dice if he got five sixes and a two. Indeed, itwould be even more appropriate to say that Tom almost got sixsixes if he rolled five sixes and a five.

Scripts for achievements often specify a series of landmarksthat provide a provisional metric of proximity to the outcome.Getting a wedding license, for example, is one of the last land-marks in the script for marriage. It will usually be appropriateto say of a couple that came that far but did not marry that theyalmost got married. However, although high propensity for anoutcome can be inferred from the near completion of the scriptfor that outcome, such inferences are tentative and dependenton default assumptions about the causal system. Thus, it is notcorrect to say of a tethered mountain climber who falls that heor she "almost fell to the bottom of the cliff," or even was closeto doing so, although the script for a fall to the bottom wasalmost completely satisfied. Nor will it be correct to say thatTom almost rolled six sixes if one of the dice has been altered tomake that outcome impossible. The propensity for a counter-factual outcome cannot be reduced to a superficial assessmentof the similarity of the actual episode to the completed scriptfor that outcome.

Intentions can contribute to an impression of propensity. Forexample, it is more appropriate to say that the escaping mur-derer was almost killed by a shot that went six inches above hisor her head if the shot was intended to kill than if it was in-tended to warn. Intentions do not suffice, however, when thereare significant obstacles to be overcome. For an individual to"consider doing X" is sometimes sufficient to support the infer-ence that the individual "almost did X" but not always. Selectedexamples follow:

7. Martin considered getting married to Meg. Martin almostmarried Meg.


8. Neil considered not getting married to Amanda. Neil almostdidn't marry Amanda.


9. Fred considered stealing his child's savings. Fred almost stolehis child's savings.


10. Ned considered breaking into a bank vault. Ned almost brokeinto a bank vault.


Mere consideration of a marriage is not sufficient (at least inthis culture) to support the assertion that the marriage almosttook place. The situation is somewhat different in Example 8,because either party (again in this culture) has the power single-handedly to put a stop to plans to marry. Responses to Exam-ples 9 and 10 show that subjects are sensitive to the fact thatmuch more remains to be done, beyond mere consideration, forthe project of breaking into a bank vault than for stealing one'schild's savings.

Decisiveness

Many outcomes are produced by a conjunction of events, allcontributing to making the outcome necessary. It is useful todistinguish two privileged roles of events in multiple causation:Critical events are those that initiate a causal episode, potentiatesubsequent causal events, or both; decisive events are those thatrule out all alternatives, and ensure (or almost ensure) a particu-lar outcome. The special role of critical events that initiate co-herent causal episodes has been confirmed in studies of blame(J. T. Johnson, Ogawa, Delforge, & Early, 1989) and studies ofmental simulations that "undo" outcomes (Wells et al., 1987).The person who starts a quarrel will get much of the blame forits consequences. However, the decisive and irreversible eventsthat terminate causal episodes are also important, especiallywhen the events in the causal sequence are not themselvescausally related (D. T. Miller & Gunasegaram, 1990). Hart andHonore (1959) proposed that a cause is found by "tracing back"from the effect to the nearest plausible candidate in the causalchain. They also discussed the legal doctrine of the last clearchance: The last person who had a good chance to avoid harmis alone held responsible (see also Wells & Gavanski, 1989). Theresponsible individual is the one whose actions cannot be re-versed by anyone else. The same intuition shows up in the con-text of blackjack; many players believe that the player on the

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


seventh box, who receives cards immediately prior to thedealer's draw that all players are trying to beat, determines theoutcomes for all players (Keren & Wagenaar, 1985). By the timethe cards are dealt, the sequence of cards is fixed, though un-known, and the seventh player, by refusal or acceptance of acard, decisively determines its allocation.

It is instructive to analyze decisiveness in terms of probabil-ity. Consider an urn game that ends whenever the excess of ballsof one color reaches a critical value. Suppose the prior probabil-ity of red being the "winning" color is high, because there aremore red than blue balls in the urn. Now imagine anotherscenario, which involves a balanced urn and a majority of redballs in early draws. When the objective probabilities of a redvictory are matched in these two scenarios, the probabilities oftwo more specific events will be higher in the case favored byevent cues: (a) the probability of the outcome occurring soon,and (b) the probability that the current lead will be preserveduntil the end of the game. We suggest that impressions of pro-pensity are related to the probability of the next favorable eventbeing decisive, and of current progress not being reversed be-fore the outcome is reached.

The intuition that causal impact increases in the course ofthe episode is especially compelling when the episode termi-nates at a fixed time. Obviously, the probability that a team thatleads by a touchdown will win the ball game must increase astime remaining to play diminishes. A score that changes thelead is accordingly perceived as more likely to be decisive if itcomes late rather than early in the game. Correspondingly, theclose counterfactual is most compelling if the propensity for theunrealized outcome peaked late in the causal episode. An earlyevent may support a counterfactual such as "Team A could havewon if Fred had not missed that touchdown in the firstquarter," but the description "Team A almost won" is muchmore convincing if the missed touchdown happened in theclosing minutes of the game.

Propensities for all outcomes will be weak in the early phasesof a causal episode, if no decisive advantage can be gained atthat time. Early in a football game, neither team has a strongpropensity to win, although one of them may have a strongdisposition to do so. Later on, propensities to win will be attrib-uted to a team to the extent that it already has, or appears on itsway to achieving, a lead that is likely to be maintained to theend. Toward the end of the game, a team with a large lead hasan overwhelming propensity to win, and both teams have asignificant propensity if the game is close. On the usual inter-pretation of probability, of course, the sums of the probabilitiesof victory for the two teams (barring ties) should add to one atall stages of the game. A formal representation of propensitiesshould incorporate the attribute of noncomplementarity, whichis admissible in some nonstandard models of probability(Shafer, 1976).

Competitive Causation

The psychological concept of propensity that was introducedin the preceding section has a dual meaning as a probabilisticand as a causal notion. We have interpreted propensity as anintuitive assessment of the current probability of the focal out-come based on event cues, and also as an assessment of the

current probability of particular cases of the focal outcome—for example, the event of this outcome occurring soon. But theterm propensity was chosen because it also denotes a directexpression of causal force—Webster's Dictionary defines pro-pensity as "an urgent and often intense natural inclination."Urgency and intensity are not part of the meaning of probabil-ity in theoretical discourse. We suggest, however, that thesedynamic features are important aspects of the cognitive repre-sentation of many causal processes, including, in particular, theprocesses that have achievements as outcomes.

The probabilistic and the causal aspects of propensity sug-gest different representations of the relation between the alter-native outcomes of a causal process. In the language of probabil-ity, this relation is expressed by complementarity: Changes inthe probability of the focal outcome are mirrored by compen-sating changes in the aggregate probability of other outcomes.In the language of causal dynamics, the relation between alter-native outcomes is best described as competition and conflict.The competition metaphor is evident in many phrases chosento describe episodes and their outcomes (e.g., "They had toadmit defeat and gave up hope of beating the deadline" or "TheHarvard job offer won out"). A competitive model of causationis particularly appealing for close counterfactuals, where thestrongest propensity is associated first with one outcome, thenwith another—suggesting a shifting balance between variableopposing forces.

A schema of competing and interacting propensities is mostobviously applicable to athletic contests, from which several ofour examples have been drawn, but is not restricted to thesesituations. Displays of the chance games that we have discussedinvite a competitive interpretation, much as the figures in thefamous Heider and Simmel (1944) animation evoke impres-sions of intentionality and meaningful interaction. We proposethe general hypothesis that the competitive schema is com-monly evoked by situations in which the focal outcome is anachievement. These include such varied cases as the making ofa difficult individual decision, the vicissitudes of a couple thatmay or may not break up or get married, the struggle of a firmthreatened with bankruptcy, the story of a life-threatening ill-ness, the construction of a building under time constraints, andthe wrecking of a building by a tropical storm. Each of thesesituations is defined by one or more focal achievements. Causalepisodes that produce such achievements, or fail to producethem, are naturally described as a struggle of conflicting andvariable forces favoring alternative outcomes, or in some casesas a struggle between a single variable force and a series ofobstacles.

The notion of conflict between opposing forces is not new topsychological analyses of causality, at least in the context ofexplaining action. Lewin (1936) introduced motion in a forcefield as a model of action under conflict. His theory influencedHeider's subsequent analysis of the naive theory of action, inwhich action is the resultant of the effective personal force andthe effective environmental force (Heider, 1958). Both modelsexplain action as a vectorial combination of forces. Lewin's fa-mous theory of conflict also incorporated a dynamic element:The forces acting on the individual change predictably as theindividual moves toward sources of attraction or away fromaversive states. In general, however, applications of force field

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


analysis have been static. There has been little emphasis ontime or on the possibility of causal forces interacting and chang-ing in the course of an event.

A model of competing propensities would extend Lewinianforce field analysis in several ways. First, the concept of focalachievement applies to outcomes that are not goals, such assomeone dying from a disease, and to situations that do notinvolve intentions at all, such as a storm destroying a building.Second, the focus of the competitive model is on extendedcausal episodes. Third, the competitive model attributes inher-ent uncertainty to causal systems and describes causal episodesin terms of propensities that may change and interact.

Psychology—or Philosophy?

The present study occupies a somewhat uncomfortable mid-dle ground between psychology and philosophy. We have at-tempted to identify the conditions under which a particularclass of counterfactual assertions would be considered true, orappropriate, and we have introduced a notion of propensity toaccount for these observations. The questions we addressed aresimilar to those modern philosophers often raise: Philosophi-cal analyses of counterfactuals, for example, focus on the truthconditions or assertability conditions for counterfactual condi-tionals. Furthermore, some elements of the method are similar:Persuasive philosophical arguments commonly draw on com-pelling examples that evoke strong shared intuitions. Althoughthe final product of philosophical analysis often has the form ofa formal deductive system, induction from intuitions about par-ticular examples is clearly an important part of philosophicalendeavor. However, there are important differences betweenthe aims and assumptions of the two disciplines. Philosopherstry to understand causality, probability, or counterfactual con-ditionals, whereas psychologists try to understand how peoplethink about these topics. These different aims have importantconsequences in the attitude toward logical consistency: Un-derstanding a matter involves imposing a consistent logicalstructure on it, but the study of human thinking should neitherassume nor impose consistency on its subject matter.

There is a large and interesting philosophical literature oncounterfactuals (e.g., Adams, 1976; Goodman, 1954; Lewis,1973,1979; Nute, 1980; Pollock, 1976; Skyrms, 1980; Stalnaker,1968).4 After developing our notions of causal episodes andchanging propensities, we encountered similar ideas in Kvart'streatment of counterfactuals and in his later work on causality(Kvart, 1986,1989). Kvart (1986) described the truth conditionsfor counterfactual conditionals on the basis of causal processesdiverging from actual historical processes at a particular pointin time. He introduced a notion of causal paths, explicated byreference to conditional probabilities changing over time. Kvartalso emphasized that the commonsense view of the world isnondeterministic, involving a concept of an open future. Asmight be expected in a philosophical analysis, Kvart treatedcounterfactuals as objects of thought, not as constructions ofthe mind. He also had recourse to formal notions of probabilityand to formal constraints on causal paths, which we haveavoided.

Psychologists have drawn most heavily on the tools and con-cepts of logical and philosophical analysis in studies of deduc-

tive reasoning (Braine, 1978; Johnson-Laird, 1983; Rips, 1990).The costs of such borrowing could be high in studies of causal-ity, probability, and counterfactuals. The intensity of currentphilosophical debate regarding these topics suggests the exis-tence of compelling but mutually inconsistent intuitions. Theconcepts that have been developed in attempts to resolve theseinconsistencies are sometimes quite remote from the naive cate-gories of thought with which psychologists are concerned. Justas an understanding of naive physics may benefit more fromacquaintance with Aristotelian physics than with the modernvariety, psychological studies of causality, probability, andcounterfactuals may do well to avoid exaggerated dependenceon the categories of modern philosophical thought.

The dominant approach to causality in psychology, perhapsreflecting a similar dominance in philosophy, treats causationas a particular relationship of dependency between events—ex-pressed by necessary or sufficient conditions or by increasedconditional probabilities (Einhorn & Hogarth, 1986; Kelley,1967; Mackie, 1974). There is another view, however, whichtreats causality as a directly perceived link between events or asan emergent property of a patterned sequence of events. Themain sources of this approach to causality in psychology arestill the classic works by Michotte (1946) and Heider and Sim-mel (1944), which, respectively, explored variations on thethemes of spatiotemporal contiguity and of schemas of inten-tional action. Ducasse (1969) has developed a philosophicalanalysis that draws on similar intuitions. The notion of propen-sity that has been presented here belongs to this tradition ofresearch in causality.

In our use of the term, the representation of propensity isinherently causal, and inherently predictive, much like the per-ception of an object in motion (Freyd & Finke, 1984). Our em-phasis on event cues to propensity deliberately straddled thestandard distinction between causes and effects as well as thedistinction between causal force and probability. Is there a jus-tification for a concept that blurs accepted distinctions betweenimportant categories of thought? There may be. We have de-scribed propensity as a perceived attribute with objective refer-ence, much like the perceived length of a line or the perceiveddistance of an object. Even in the case of lengths and distances,the crude correspondence of the dimensions of percepts to thedimensions of physical description of the world does not guar-antee correspondence of the geometries that describe the spacepeople perceive and the space in which they move. The moregeneral point is that the mental representations of events andtheir relations may not correspond to any logical analysis ofcausality or probability, and that intuitions about these mattersmay not be internally consistent. The student of lay intuitionsfaces a problem that is familiar to cultural anthropologists:How does one make sense of a system of thought without im-posing alien categories on it?

4 Skyrms (1980) has a treatment of counterfactual conditionals thatrelies on what he calls "prior propensities." However, the meaning ofhis term more closely resembles our usage of dispositions.

References

Adams, E. W (1976). Prior probabilities and counterfactual condition-als. In W L. Harper & C. A. Hooker (Eds.), Foundations of probability

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


theory, statistical inference, and statistical theories of science (Vol. 1,pp. 1-21). Dordrecht, Holland: Reidel.

Ajzen, I. (1977). Intuitive theories of events and the effects of base-rateinformation on prediction. Journal of 'Personality and Social Psychol-ogy, 55,303-314.

Bar-Hillel, M. (1990). Back to base rates. In R. M. Hogarth (Ed.), In-sights in decision making: A tribute to Hillel J. Einhom (pp. 200-216).Chicago: University of Chicago Press.

Black, J. B., & Bower, G. H. (1979). Episodes as chunks in narrativememory. Journal of Verbal Learning and Verbal Behavior, 18, 187-198.

Braine, M. D. S. (1978). On the relation between the natural logic ofreasoning and standard logic. Psychological Review, 85,1-21.

Ducasse, C. J. (1969). Causation and the types of necessity. New \fork:Dover.

Einhorn, H., & Hogarth, R. (1986). Judging probable cause. Psychologi-cal Bulletin, 99,3-19.

Fielding, H. (1975). The History of Tom Jones, a foundling (Vol. 2).Oxford, England: Wesleyan University Press/Oxford UniversityPress. (Original work published 1749)

Fischhoff, B. (1975). Hindsight i= foresight: The effect of outcomeknowledge on judgment under uncertainty. Journal of ExperimentalPsychology: Human Perception and Performance, 1, 288-299.

Fischhoff, B. (1982). For those condemned to study the past. In D.Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncer-tainty: Heuristics and biases (pp. 422-444). Cambridge, England:Cambridge University Press.

Freyd, J., & Finke, R. (1984). Representational momentum. Journal ofExperimental Psychology: Learning, Memory and Cognition, 10,126-132.

Gigerenzer, G., Hell, W, & Blank, H. (1988). Presentation and content:The use of base rates as a continuous variable. Journal of Experimen-tal Psychology: Human Perception and Performance, 14,513-525.

Gilovich, T, Vallone, R., & Tversky, A. (1985). The hot hand in basket-ball: On the misperception of random sequences. Cognitive Psychol-ogy, 17,295-314.

Goodman, N. (1954). Fact, fiction and forecast (1st ed.). London: Ath-lone Press.

Goodman, N. (1983). Fact, fiction and forecast (4th ed.). Cambridge,MA: Harvard University Press.

Hart, H. L. A., & Honore, A. M. (1959). Causation in the law. London:Oxford University Press.

Heider, F. (1958). The psychology of interpersonal relations. New \brk:Wiley.

Heider, F., & Simmel, M. (1944). An experimental study of apparentbehavior. American Journal of Psychology, 57, 243-259.

Hofstadter, D. (1979). Godel, Escher, Bach: An eternal golden braid.New \fork: Basic Books.

Johnson, J. T. (1986). The knowledge of what might have been: Affec-tive and attributional consequences of near outcomes. Personalityand Social Psychology Bulletin, 12, 51-62.

Johnson, J. T, Ogawa, K. H, Delforge, A, & Early, D. (1989). Causalprimacy and comparative fault: The effect of position in a causalchain on judgments of legal responsibility. Personality and SocialPsychology, 15,161-174.

Johnson, M. K. (1988). Discriminating the origin of information. InT. F. Oltmanns & B. A. Maher (Eds.), Delusional beliefs: Interdisci-plinary perspectives (pp. 34-65). New York: Wiley.

Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive scienceof language, inference, and consciousness. Cambridge, MA: HarvardUniversity Press.

Kahneman, D., & Miller, D. T. (1986). Norm theory: Comparing realityto its alternatives. Psychological Review, 93,136-153.

Kahneman, D, Slovic, P., & Tversky, A. (Eds.). (1982). Judgment under

uncertainty: Heuristics and biases. New York: Cambridge UniversityPress.

Kahneman, D., & Tversky, A. (1973). On the psychology of prediction.Psychological Review, 80, 237-251.

Kahneman, D., & Tversky, A. (1982a). The simulation heuristic. In D.Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncer-tainty: Heuristics and biases (pp. 201-208). New York: CambridgeUniversity Press.

Kahneman, D., & Tversky, A. (1982b). Variants of uncertainty. Cogni-tion, 11,143-157.

Kelley, H. H. (1967). Attribution theory in social psychology. In D.Levine (Ed.), Nebraska Symposium on Motivation (Vol. 15, pp. 192-240). Lincoln: University of Nebraska Press.

Keren, G., & Wagenaar, W A. (1985). On the psychology of playingblackjack: Normative and descriptive considerations with implica-tions for decision theory. Journal of Experimental Psychology: Gen-eral, 114,133-158.

Kintsch, W, & van Dijk, T. A. (1978). Toward a model of text compre-hension and production. Psychological Review, 85, 363-384.

Kvart, I. (1986). A theory of counterfactuals. Indianapolis: Hackett.Kvart, I. (1989). Causal independence. Unpublished manuscript.Lakoff, G. (1987). Women, fire, and dangerous things: What categories

reveal about the mind. Chicago: University of Chicago Press.Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago: Uni-

versity of Chicago Press.Landman, J. (1987). Regret and elation following action and inaction.

Personality and Social Psychology Bulletin, 13, 524-536.Lewin, K. (1936). Principles of topological psychology (F. Heider &

G. M. Heider, Trans.). New York: McGraw-Hill.Lewis, D. (1973). Counterfactuals. Cambridge, MA: Harvard University

Press.Lewis, D. (1979). Counterfactual dependence and time's arrow. Nous,

13, 455-476.Lyons, J. (1977). Semantics, Volume 2. Cambridge, England: Cambridge

University Press.Mackie, J. L. (1974). The cement of the universe: A study of causation.

Oxford, England: Clarendon Press (Oxford University Press).Michotte, A. E. (1946). La perception de la causalite [The perception

of causality]. Louvain, Belgium: University of Louvain Publica-tions.

Miller, D. T, & Gunasegaram, S. (1990). Temporal order and the per-ceived mutability of events: Implications for blame assignment.Journal of Personality and Social Psychology, 59,1111 -1118.

Miller, D. T, Turnbull, W, & McFarland, C. (1990). Counterfactualthinking and social perception: Thinking about what might havebeen. In M. Zanna(Ed.), Advances in experimental social psychology(Vol. 23, pp. 305-331). Orlando, FL: Academic Press.

Miller, G. A., & Johnson-Laird, P. L. (1976). Language and perception.Cambridge, England: Cambridge University Press.

Nute, D. (1980). Topics in conditional logic. Boston: Reidel.Pennington, N., & Hastie, R. (1988). Explanation-based decision mak-

ing: Effects of memory structure on judgment. Journal of Experi-mental Psychology: Learning, Memory and Cognition, 14, 521-533.

Pollock, J. L. (1976). Subjunctive reasoning. Dordrecht, Holland: Rei-del.

Read, S. J. (1987). Constructing causal scenarios: A knowledge struc-ture approach to causal reasoning. Journal of Personality and SocialPsychology, 52, 288-302.

Rips, L. (1990). Reasoning. Annual Review of Psychology, 41,321-353.Rumelhart, D. E. (1977). Understanding and summarizing brief sto-

ries. In D. Laberge & J. Samuels (Eds.), Basic processes in reading,perception and comprehension (pp. 265-303). Hillsdale, NJ: Erl-baum.

Schank, R. C. (1975). The structure of episodes in memory. In D. G.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Bobrow & A. Collins (Eds.), Representation and understanding: Stud-ies in cognitive science (pp. 237-272). New York: Academic Press.

Schank, R. C, & Abelson, R. P. (1977). Scripts, plans, goals, and under-standing. Hillsdale, NJ: Erlbaum.

Shafer, G. (1976). A mathematical theory of evidence. Princeton, NJ:Princeton University Press.

Skyrms, B. (1980). The prior propensity account of subjunctive condi-tionals. In W L. Harper, R. Stalnaker, & G. Pearce (Eds.), Ifs: Condi-tionals, belief decision, chance and time (Vol. 1, pp. 125-162). Dor-drecht, Holland: Reidel.

Stalnaker, R. C. (1968). A theory of conditionals. In N. Rescher (Ed.),Studies in logical theory: American Philosophical Quarterly mono-graph series 2 (pp. 98-112). Oxford, England: Blackwell.

Talmy, L. (1981, May). Force images. Paper presented at the Conferenceon Language and Mental Imagery, University of California, Berke-ley.

Talmy, L. (1983). How language structures space. In H. Pick & L. Acre-dolo (Eds.), Spatial orientation: Theory, research and application (pp.219-282). New York: Plenum Press.

Trabasso, T., Secco, T, & van den Broek, P. (1984). Causal cohesion and

story coherence. In H. Mandl, N. L. Stein, & T. Trabasso (Eds.),Learning and comprehension of text (pp. 83-111). Hillsdale, NJ: Erl-baum.

Tversky, A., & Kahneman, D. (1980). Causal schemas in judgmentsunder uncertainty. In M. Fishbein (Ed.), Progress in social psychol-ogy (Vol. 1, pp. 153-160). Hillsdale, NJ: Erlbaum.

Tversky, A., & Kahneman, D. (1982). Evidential impact of base rates. InD. Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncer-tainty: Heuristics and biases (pp. 201-208). New York: CambridgeUniversity Press.

Vendler, Z. (1967). Linguistics in philosophy. Ithaca, NY: Cornell Uni-versity Press.

Wells, G. L., & Gavanski, I. (1989). Mental simulation of causality.Journal of Personality and Social Psychology, 56,161-169.

Wells, G. L., Taylor, B. R., & Turtle, J. W (1987). The undoing of sce-narios. Journal of Personality and Social Psychology, 53, 421-430.

Received December 30,1989Revision received July 5,1990

Accepted July 9,1990 •

Harvey Appointed Editor of Contemporary Psychology, 1992-1997

The Publications and Communications Board of the American Psychological Associationannounces the appointment of John H. Harvey, University of Iowa, as editor of ContemporaryPsychology for a 6-year term beginning in 1992.

Publishers should note that books should not be sent to Harvey. Publishers should continue tosend two copies of books to be considered for review plus notices of publication to

PsycINFO Services Department, APAAttn: Contemporary Psychology Processing1400 North Uhle StreetArlington, Virginia 22201

Please note that all reviews are written by invitation.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

propensities and counterfactuals: the loser that almost won 0lli... · attitudes and social...

Documents